Francesca | Data science PhD
Region: GB
Friday 27 June 2025 17:12:06 GMT
168153
19656
130
645
Music
Download
Comments
leonrdo :
For AI models to reason, I truly believe we need to go back to the drawing board. I struggle to believe our current models can genuinely think. But who knows, OpenAI are generally close source so perhaps they’re making something mind blowing
2025-06-27 23:46:14
334
pinkksatan :
where do u get the recommendations of papers? or do you just pick them based on your interests?
2025-07-01 06:46:13
0
Lucrece Borgia :
OK interested in the linear regressions for non normal data....
2025-06-28 10:13:08
92
Codecademy :
wait, what a fun roundup! thanks for sharing!
2025-06-30 21:05:04
6
MeliMel :
How do I get access to these papers?
2025-06-30 20:43:42
1
L.J. :
I thought it was always the residuals that had to be normal, not the actual outcome?
2025-07-01 16:14:26
0
Lolita Torres :
i read thing like this for fun & wonder why i cant relate to the ppl around me :/
2025-07-01 17:35:31
28
jack :
Hi! How do you find interesting/important papers to read?
2025-07-01 09:09:50
0
Lamb :
Bro that second slide is sick. Always thoughts transformations (like finding exponents which minimize RSS) was weird and p-hacky. Looking forward to reading that paper.
2025-06-30 05:07:17
0
wittgenfine :
The ghost of my first-year stats professor hovers besides me, whispering in my ear: “it’s the residuals that should be normally distributed”
2025-06-30 11:52:52
0
Penelope :
Wow linear regression was so interesting
2025-06-28 08:50:32
6
Max :
sad that the flawed Apple paper still get so much traction
2025-07-01 20:45:24
0
Prvsha :
I wanna get back into data science I’ve gone down the Investment route . Read itslwr 3 years ago, still fiddle with python . But fr how do I regain my knowledge
2025-06-28 09:49:42
4
shinr4_ :
I wanna read the Linear Regression and normality assumptions, may I have the link to the paper?
2025-06-30 08:08:29
0
BM :
For linear regression you just need the residuals to be normal iid
2025-06-28 09:04:32
1
Silot :
Why are we so fixated on reasoning? We can reason. Let’s use LLMs to become better at reasoning
2025-06-28 07:50:59
139
Katie :
Going to have to read that linear regression model paper
2025-06-30 01:32:43
39
Cristofer Franco Vil :
fun fact is LLM struggle to answer easy questions (8 disk Tower of Hanoi) even with PB of data!! isn't that weird? we can do better even with much more less data
2025-06-29 05:47:26
9
Cristian Ruz Sanchez :
u should read the ilusión of ilusión of thinking
2025-06-27 21:01:35
7
Tom Zeeuwe :
Replying to get more of this on my fyp
2025-06-28 11:29:09
7
M :
I’m going to read that regression & normality paper later. Thank you for sharing.
2025-06-29 20:34:23
6
Lorenzo Pla :
I HAVE FOUND MY PEOPLE ❤️
2025-06-30 11:22:22
5
Hectic Content :
I want to stay on this side of tiktok. Unrotting our brains!!
2025-09-12 14:30:30
4
aditigro_ :
You seem like a really cool person to have a late night talk with
2025-06-28 17:49:44
3
The Fragrance Sampler :
The one by Apple is incredibly biased from what I’ve heard. I haven’t actually read it, but the limits that they conclude as an inability to reason seem to actually be due to the strict token limit they imposed and the fact that it didn’t let the models use code, but if the limit is increased and code is used (as shown by other studies) the results are vastly different pointing to the ability to reason. The bias and conflict of interest I’m referring to comes from Apple’s poor performance in the AI marketplace, so it makes sense that they’d publish papers denouncing the ability of LLMs and other reasoning models as a whole. To respond to your point specifically, this would mean that ChatGPT models, which can use code and have a high token limit, may actually have the ability to reason.
2025-06-28 00:24:15
3
Lolly :
I wish I was this productive 😭😭😭
2025-06-28 11:37:30
2
KhanyisaMzima :
commenting to stay on this side of tiktok
2025-08-01 15:33:32
2
Fisherman :
I love it here
2025-07-29 15:13:01
2
Natalie Flora :
I feel like workflows can be implemented to replicate an illusion of reasoning, but it will always be that — an illusory house of cards, delicately constructed and tuned
2025-06-29 22:17:15
2
Itchymitchy :
Ty, commenting for more of this
2025-07-01 18:37:45
1
Lena :
LLM is not the same thing as AI and yet the two have become synonymous. Of course an LLM can’t reason, it can only retrieve data like a database
2025-07-01 12:55:55
1
Manu Pasta :
Thanks for the list 🥰
2025-07-24 03:32:12
1
⛺ Founder building Campfire :
On Illusion of thinking: the paper doesn’t prove LRMs/LLMs can’t reason at all, but it does show that current incarnations break down sharply when complexity exceeds certain thresholds, suggesting these models aren’t simply scaling like digital mathematicians (they overthink easy problems and fall apart with hard ones). The opportunity still exists to improve these models, net it’s not so much a can’t, more like not yet.
2025-06-30 16:23:35
1
Sandy deephud :
this is super cool!
2025-07-07 17:48:40
1
ensofoto :
love this <3
2025-06-30 13:46:14
1
w1nt3rmut0 :
That last paper 🫡 Hernan’s Causal Inference is a great book
2025-07-01 02:41:09
1
DelinquentWh0re :
Left to go wash dishes and completely forgot about this and let this play for a whole 24 minutes. That’s standing by girls.❤️🙏
2025-07-01 18:44:54
1
Semut Korban :
can i get the link for all those article please
2025-06-30 08:18:48
1
Shakedguyer :
Where can we read????😁🥰😊😌
2025-08-20 17:11:35
1
{Human}{brand}{Artist} :
Thanks !
2025-06-28 09:07:08
1
es :
instant follow wow
2025-09-02 19:53:32
1
João Carvalho :
I'm not sure about the interpretation of the reasoning paper, I'm not in favor of the big AI hype, but interacting with reasoning models makes me wonder if thinking really is needed, or that if they can't think maybe we can't either
2025-06-29 15:36:40
1
Buntler :
I don’t understand why researchers are so into kneecapping the AI by removing its ability to write+execute code, then claiming it “can’t reason”. how is that productive?
2025-07-02 06:31:45
0
Joey So :
🫶🏿
2025-07-06 06:13:14
0
NorthSouth :
✨👍
2025-07-02 02:37:13
0
Basalty | Geology + Whatever :
oh I need that normality assumption paper
2025-07-02 01:46:22
0
Mymy :
Immediate follow
2025-07-01 22:01:27
0
player1857 :
lol slide 3 i believed it before looking at your post. not a genius, actually just didnt care
2025-07-01 21:03:48
0
prince_toks :
I would love to hear your take on these articles
2025-07-01 20:18:13
0
Pax | AI IRL :
This are so interesting. I’m getting a masters in AI implementation. I’m going to have to dig in to these
2025-09-19 12:26:30
0
Piscapalion :
structural models the woat
2025-07-02 14:43:52
0
sukhyun_127 :
Why did you think you needed normally distributed outcomes for regression...?
2025-07-03 15:42:10
0
AIburbclave :
Casual modeling 👏
2025-08-01 20:10:31
0
si machu pa :
why I use A.I for coding and grammar checker. 😅
2025-07-07 09:01:05
0
To see more videos from user @frantalkstech, please go to the Tikwm
homepage.
